Best way to look at this is through examples: Companies like FaceBook or eBay or name-your-cloud-service-provider have designed their multi-tenant services around a certain kind workload can that scale like crazy, built on a highly integrated (and often propriety) infrastructure stack that is designed push resiliency up into the software layer. This lets them use really inexpensive, commoditized components that can fail without bringing the whole service down. And, the multi-tenant software lets them avoid the costs associated with running multiple software instances (the ASP model).

Contrast this with a typical enterprise, let’s say a multi-national manufacturing concern, who needs to execute a multitude of different workload types (think: supply chain, ERP, collaboration, etc) mostly inside their firewall, but of course accommodating supply chain partners, customer etc. It would be smart for this company to virtualize and automate their servers/storage/network as much as possible, consolidate app instances and the like. It would also be smart to provide their internal infrastructure as a service, so that they can efficiently provision configure-to-order infrastructure for their internal constitutions. They might also choose to use outsourced or hosting partners for this internally-consumed infrastructure.

But by no means would they construct the same kind of single-purpose, massive scale out, global-class infrastructure as the big cloud service companies. It’s apples and oranges.

The market wants to call both of these things “cloud” which causes the confusion.

So, HP has settled on distinguishing global-class services (you might call it cloud-class services) from enterprise-class services. Both are necessary, both serve different purposes, both will live on.

I think you’ll find the first virtual machines and hypervisors were first provided by IBM and having more than 40 years of insight into virtualisation might be worth consideration.

http://en.wikipedia.org/wiki/Hypervisor

Unless you use shared and off-premise (in HP terminology) clouds, You are not going to really rack up the cost saving or any of the advantages of true “cloud computing”

Dedicated and on premise clouds is known by another name: “Virtualization” . Dont go to HP or IBM for that solution, go directly to the source which is VMware, Microsoft or Xen. They will happily sell you virtualisation software and tools to manage them …

On premise shared clouds are hard to implement. Who are you going to share the data centre with and why is this any more “secure” than hosting your application on Amazon in the first place ??

The only market with promise which has not fully been attacked is the off premise dedicated clouds. This is essentially the same as creating a cloud with a firewall in it blanketing everybody else from those machines … For cloud vendors like Amazon, Azure – Implementing that technology will be key to getting Enterprises on your platform anyway …